alternative to statsby that gives significance levels

[Stan Peterburgsky]

Hello,
My understanding is that statsby ... : regress ... doesn't give significance levels. Is there an alternative that provides significance levels (either in the form of stars or p-values)?
Thank you,
Stan

[Clyde Schechter]

if you specify _b _se as the list of expressions for -statsby- to save, you can calculate your own pvalues from them. This example, using the auto.dta illustrates the approach:

clear*
sysuse auto

tempfile results
statsby _b _se e(df_r), by(rep78) saving(`results'): regress price mpg headroom

use `results', clear
foreach x in mpg headroom cons {
    gen t_`x' = _b_`x'/_se_`x'
    gen p_`x' = 2*ttail(_eq2_stat_1, abs(t_`x'))
}

[Stan Peterburgsky]

Great, thank you very much! Didn't know how to extract degrees of freedom.